I have CPDN and S@H running in a stipulated 1:2 ratio across 2 processors, and have preferences set at "keep 7 days workload", because there are times when I'm away from a net connection with my laptop for a few days at a time. What I find is that the ration of average work done ranges from about 1:5 to about 1:1. typically the S@H workunits all get executed, such that the avg work done is about 1:5, and when BOINC runs out of S@H workunits, it determines that it needs to give CPDN a go for a while and does not acquire any more S@H work. The ratio of avg work done gradually rises to about 1:1.5 or sometimes nearly to 1:1, then BOINC realises it needs a bunch of S@H and so acquires a week's worth, all with deadlines of a week or so away. It then decides that it is going to have trouble doing them all in a week and starts running them all with priority which means that for the next week S@H runs almost exclusively and CPDN gets almost nothing. So at the end of the week, there is no S@H work left and the ratio is back to about 1:5, so BOINC runs CPDN and doesn't ask for any more S@H, and so we go again.

This boom and bust cycle plays out endlessly. How do I get BOINC to keep a week's worth of S@H but still get work done progressively?

---

What are your current setting for
Computer is connected to the Internet about every
and
Maintain enough work for an additional

For the way you operate I think that the best setting would probably be connection 3 and additional 4.

It is normal for any WU with only 7 day deadline to be rushed ahead of other work (priority) since Boinc knows that they must be ready to upload before your next connection time.

edit: Boinc priority number one is to make sure WUs are returned before deadline date. It then give the other project extra time so as to maintain your resource share in the long term.

---

Boinc V7.2.42
Win7 i5 3.33G 4GB, GTX470

I see the issue with the 7 day connect interval, and have changed it to 3 + 4. But I don't think that's the main problem. The last batch of workunits I got was about a week ago, and they all came with a deadline of 23 or 24 Nov, and 45 hours of computing required, which given the number of WUs amounted to about as many hours as there were between arrival and deadlines. It's as though it obtained 21 days of computing when I asked to always have 7 available. BOINC's enthusiasm is a little excessive. My usage patterns are fairly consistent, so BOINC can't use the excuse that it thought it could get through them all. As soon as they arrived (like, instantly) S@H went into high priority mode. Which is a paradox.

---

Because you are away from the internet for a week at a time, the correct setting for "Connect every X" is 7 days. Because this conflicts with some of the S@H deadlines, you are just going to have the boom and bust cycles.

What happens is that the long term debt for S@H drops below the cuttoff for work fetch while it is in exclusive mode. Then CPDN runs for a while while the Long Term Debt of S@H slowly rises. Once the LTD of S@H has risen enough, it gets more work.

The only way to break this is to allow connection to the internet more frequently, and cut the "Connect every X" setting appropriately.

Please note that the "Connect Every X" setting is used to ensure that work is completed enough before the report deadline so that it can be reported on time - even though the computer may not be connected for the interval specified by "Connect Every X". Setting this value lower than the actual expected time offline runs a substantial risk of returning work late and losing all credit for that work as it may be scientifically useless by the time it gets returned.

---

BOINC WIKI